An Arabic-Native NLP Conversational Framework for Intelligent Academic Advising and Major Selection | ||
| المجلة الدولية للتكنولوجيا والحوسبة التعليمية | ||
| Article 3, Volume 4, Issue 13, October 2025 | ||
| Document Type: الدراسات والبحوث العلمية. | ||
| DOI: 10.21608/ijtec.2025.424546.1013 | ||
| Authors | ||
| S. M.khater* ; M. Hussin | ||
| Faculty of specific education computer science mansoura university | ||
| Abstract | ||
| First-year students in multidisciplinary faculties often struggle to select majors that align with their interests. This challenge is amplified in Arabic contexts due to the limited number of NLP resources for Modern Standard Arabic (MSA) and its dialects. This study presents ARAB_NLP, a conversational system integrating dialect normalization, intent classification, Arabic-adapted keyword extraction, sentiment analysis, and a knowledge base–driven retrieval and ranking module. A custom prompt-engineering framework ensures context-aware and culturally relevant recommendations delivered fully in Arabic. This study involves 50 students at Mansoura University who have strong performance: an average satisfaction score of 4.48/5 (≈90%), and high ratings for clarity (4.8), usability (4.7), and recommendation accuracy (4.6). Comparison with human advisors confirmed 86% exact matches, 10% partial, and only 4% unmatched, 96% overall alignment. Model-level evaluation further confirmed robust performance, with 92% dialect normalization accuracy, 89% F1 for intent detection, 84% Precision @5 for keyword extraction, and 91% Recall@5 for knowledge retrieval. These results demonstrate that an Arabic-first, knowledge base–driven NLP pipeline can provide accurate and culturally relevant academic advising for first-year students. | ||
| Keywords | ||
| Academic Advising; NLP; First-Year Students; MSA. Retrieval-Augmented Generation (RAG); Department Recommendation | ||
|
Statistics Article View: 92 |
||
